# Research Papers — 2026-03 · bundle page 115 of 243 > Full-text markdown of papers #5701–5750 of 12123 published in 2026-03, newest first. AI-readable mirrors of the full 12123-paper 2026-03 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-03/page/115 [← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/114.md) [Next bundle →](https://aipod.dev/papers/archive/2026-03/page/116.md) ### Robust Generative Audio Quality Assessment: Disentangling Quality from Spurious Correlations Authors: Kuan-Tang Huang, Chien-Chun Wang, Cheng-Yeh Yang, Hung-Shin Lee, Hsin-Min Wang, Berlin Chen Published: 2026-03-17 Categories: eess.AS, cs.AI, cs.SD, eess.SP arXiv: https://arxiv.org/abs/2603.16201v1 The rapid proliferation of AI-Generated Content (AIGC) has necessitated robust metrics for perceptual quality assessment. However, automatic Mean Opinion Score (MOS) prediction models are often compromised by data scarcity, predisposing them to learn spurious correlations-- such as dataset-specific acoustic signatures-- rather than generalized quality features. To address this, we leverage domain adversarial training (DAT) to disentangle true quality perception from these nuisance factors. Unlike prior works that rely on static domain priors, we systematically investigate domain definition strategies ranging from explicit metadata-driven labels to implicit data-driven clusters. Our findings reveal that there is no "one-size-fits-all" domain definition; instead, the optimal strategy is highly dependent on the specific MOS aspect being evaluated. Experimental results demonstrate that our aspect-specific domain strategy effectively mitigates acoustic biases, significantly improving correlation with human ratings and achieving superior generalization on unseen generative scenarios. --- ### Online Semi-infinite Linear Programming: Efficient Algorithms via Function Approximation Authors: Yiming Zong, Jiashuo Jiang Published: 2026-03-17 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.16200v1 We consider the dynamic resource allocation problem where the decision space is finite-dimensional, yet the solution must satisfy a large or even infinite number of constraints revealed via streaming data or oracle feedback. We model this challenge as an Online Semi-infinite Linear Programming (OSILP) problem and develop a novel LP formulation to solve it approximately. Specifically, we employ function approximation to reduce the number of constraints to a constant $q$. This addresses a key limitation of traditional online LP algorithms, whose regret bounds typically depend on the number of constraints, leading to poor performance in this setting. We propose a dual-based algorithm to solve our new formulation, which offers broad applicability through the selection of appropriate potential functions. We analyze this algorithm under two classical input models-stochastic input and random permutation-establishing regret bounds of $O(q\sqrt{T})$ and $O\left(\left(q+q\log{T})\sqrt{T}\right)\right)$ respectively. Note that both regret bounds are independent of the number of constraints, which demonstrates the potential of our approach to handle a large or infinite number of constraints. Furthermore, we investigate the potential to improve upon the $O(q\sqrt{T})$ regret and propose a two-stage algorithm, achieving $O(q\log{T} + q/ε)$ regret under more stringent assumptions. We also extend our algorithms to the general function setting. A series of experiments validates that our algorithms outperform existing methods when confronted with a large number of constraints. --- ### Are Large Language Models Truly Smarter Than Humans? Authors: Eshwar Reddy M, Sourav Karmakar Published: 2026-03-17 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2603.16197v1 Public leaderboards increasingly suggest that large language models (LLMs) surpass human experts on benchmarks spanning academic knowledge, law, and programming. Yet most benchmarks are fully public, their questions widely mirrored across the internet, creating systematic risk that models were trained on the very data used to evaluate them. This paper presents three complementary experiments forming a rigorous multi-method contamination audit of six frontier LLMs: GPT-4o, GPT-4o-mini, DeepSeek-R1, DeepSeek-V3, Llama-3.3-70B, and Qwen3-235B. Experiment 1 applies a lexical contamination detection pipeline to 513 MMLU questions across all 57 subjects, finding an overall contamination rate of 13.8% (18.1% in STEM, up to 66.7% in Philosophy) and estimated performance gains of +0.030 to +0.054 accuracy points by category. Experiment 2 applies a paraphrase and indirect-reference diagnostic to 100 MMLU questions, finding accuracy drops by an average of 7.0 percentage points under indirect reference, rising to 19.8 pp in both Law and Ethics. Experiment 3 applies TS-Guessing behavioral probes to all 513 questions and all six models, finding that 72.5% trigger memorization signals far above chance, with DeepSeek-R1 displaying a distributed memorization signature (76.6% partial reconstruction, 0% verbatim recall) that explains its anomalous Experiment 2 profile. All three experiments converge on the same contamination ranking: STEM > Professional > Social Sciences > Humanities. --- ### S-VAM: Shortcut Video-Action Model by Self-Distilling Geometric and Semantic Foresight Authors: Haodong Yan, Zhide Zhong, Jiaguan Zhu, Junjie He, Weilin Yuan, Wenxuan Song, Xin Gong, Yingjie Cai, Guanyi Zhao, Xu Yan, Bingbing Liu, Ying-Cong Chen et al. Published: 2026-03-17 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2603.16195v2 Video action models (VAMs) have emerged as a promising paradigm for robot learning, owing to their powerful visual foresight for complex manipulation tasks. However, current VAMs, typically relying on either slow multi-step video generation or noisy one-step feature extraction, cannot simultaneously guarantee real-time inference and high-fidelity foresight. To address this limitation, we propose S-VAM, a shortcut video-action model that foresees coherent geometric and semantic representations via a single forward pass. Serving as a stable blueprint, these foreseen representations significantly simplify the action prediction. To enable this efficient shortcut, we introduce a novel self-distillation strategy that condenses structured generative priors of multi-step denoising into one-step inference. Specifically, vision foundation model (VFM) representations extracted from the diffusion model's own multi-step generated videos provide teacher targets. Lightweight decouplers, as students, learn to directly map noisy one-step features to these targets. Extensive experiments in simulation and the real world demonstrate that our S-VAM outperforms state-of-the-art methods, enabling efficient and precise manipulation in complex environments. Our project page is https://haodong-yan.github.io/S-VAM/ --- ### Structured Semantic Cloaking for Jailbreak Attacks on Large Language Models Authors: Xiaobing Sun, Perry Lam, Shaohua Li, Zizhou Wang, Rick Siow Mong Goh, Yong Liu, Liangli Zhen Published: 2026-03-17 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.16192v1 Modern LLMs employ safety mechanisms that extend beyond surface-level input filtering to latent semantic representations and generation-time reasoning, enabling them to recover obfuscated malicious intent during inference and refuse accordingly, and rendering many surface-level obfuscation jailbreak attacks ineffective. We propose Structured Semantic Cloaking (S2C), a novel multi-dimensional jailbreak attack framework that manipulates how malicious semantic intent is reconstructed during model inference. S2C strategically distributes and reshapes semantic cues such that full intent consolidation requires multi-step inference and long-range co-reference resolution within deeper latent representations. The framework comprises three complementary mechanisms: (1) Contextual Reframing, which embeds the request within a plausible high-stakes scenario to bias the model toward compliance; (2) Content Fragmentation, which disperses the semantic signature of the request across disjoint prompt segments; and (3) Clue-Guided Camouflage, which disguises residual semantic cues while embedding recoverable markers that guide output generation. By delaying and restructuring semantic consolidation, S2C degrades safety triggers that depend on coherent or explicitly reconstructed malicious intent at decoding time, while preserving sufficient instruction recoverability for functional output generation. We evaluate S2C across multiple open-source and proprietary LLMs using HarmBench and JBB-Behaviors, where it improves Attack Success Rate (ASR) by 12.4% and 9.7%, respectively, over the current SOTA. Notably, S2C achieves substantial gains on GPT-5-mini, outperforming the strongest baseline by 26% on JBB-Behaviors. We also analyse which combinations perform best against broad families of models, and characterise the trade-off between the extent of obfuscation versus input recoverability on jailbreak success. --- ### Reliable Reasoning in SVG-LLMs via Multi-Task Multi-Reward Reinforcement Learning Authors: Haomin Wang, Qi Wei, Qianli Ma, Shengyuan Ding, Jinhui Yin, Kai Chen, Hongjie Zhang Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16189v1 With the rapid advancement of vision-language models, an increasing number of studies have explored their potential for SVG generation tasks. Although existing approaches improve performance by constructing large-scale SVG datasets and introducing SVG-specific tokens, they still suffer from limited generalization, redundant paths in code outputs, and a lack of explicit reasoning. In this work, we present CTRL-S (Chain-of-Thought Reinforcement Learning for SVG), a unified framework that introduces a chain-of-thought mechanism to explicitly expose the model's reasoning process during SVG generation. To support this structured reasoning, we construct SVG-Sophia, a high-quality dataset containing 145K samples across SVG code refinement, Text-to-SVG, and Image-to-SVG tasks. By training the model to generate group-level structured SVG code, CTRL-S significantly improves structural coherence and visual fidelity. Furthermore, we adopt the GRPO algorithm and design a multi-reward optimization framework, incorporating DINO, image-text similarity, format, and code efficiency rewards. Through joint multi-reward optimization and multi-task training, our approach systematically enhances overall generation capabilities. Extensive experiments show that CTRL-S outperforms existing methods, achieving higher task success rates, superior SVG code quality, and exceptional visual fidelity. --- ### ECHO: Edge-Cloud Humanoid Orchestration for Language-to-Motion Control Authors: Haozhe Jia, Jianfei Song, Yuan Zhang, Honglei Jin, Youcheng Fan, Wenshuo Chen, Wei Zhang, Yutao Yue Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16188v1 We present ECHO, an edge--cloud framework for language-driven whole-body control of humanoid robots. A cloud-hosted diffusion-based text-to-motion generator synthesizes motion references from natural language instructions, while an edge-deployed reinforcement-learning tracker executes them in closed loop on the robot. The two modules are bridged by a compact, robot-native 38-dimensional motion representation that encodes joint angles, root planar velocity, root height, and a continuous 6D root orientation per frame, eliminating inference-time retargeting from human body models and remaining directly compatible with low-level PD control. The generator adopts a 1D convolutional UNet with cross-attention conditioned on CLIP-encoded text features; at inference, DDIM sampling with 10 denoising steps and classifier-free guidance produces motion sequences in approximately one second on a cloud GPU. The tracker follows a Teacher--Student paradigm: a privileged teacher policy is distilled into a lightweight student equipped with an evidential adaptation module for sim-to-real transfer, further strengthened by morphological symmetry constraints and domain randomization. An autonomous fall recovery mechanism detects falls via onboard IMU readings and retrieves recovery trajectories from a pre-built motion library. We evaluate ECHO on a retargeted HumanML3D benchmark, where it achieves strong generation quality (FID 0.029, R-Precision Top-1 0.686) under a unified robot-domain evaluator, while maintaining high motion safety and trajectory consistency. Real-world experiments on a Unitree G1 humanoid demonstrate stable execution of diverse text commands with zero hardware fine-tuning. --- ### Sample-Efficient Adaptation of Drug-Response Models to Patient Tumors under Strong Biological Domain Shift Authors: Camille Jimenez Cortes, Philippe Lalanda, German Vega Published: 2026-03-17 Categories: cs.LG, cs.AI, q-bio.QM arXiv: https://arxiv.org/abs/2603.16185v1 Predicting drug response in patients from preclinical data remains a major challenge in precision oncology due to the substantial biological gap between in vitro cell lines and patient tumors. Rather than aiming to improve absolute in vitro prediction accuracy, this work examines whether explicitly separating representation learning from task supervision enables more sample-efficient adaptation of drug-response models to patient data under strong biological domain shift. We propose a staged transfer-learning framework in which cellular and drug representations are first learned independently from large collections of unlabeled pharmacogenomic data using autoencoder-based representation learning. These representations are then aligned with drug-response labels on cell-line data and subsequently adapted to patient tumors using few-shot supervision. Through a systematic evaluation spanning in-domain, cross-dataset, and patient-level settings, we show that unsupervised pretraining provides limited benefit when source and target domains overlap substantially, but yields clear gains when adapting to patient tumors with very limited labeled data. In particular, the proposed framework achieves faster performance improvements during few-shot patient-level adaptation while maintaining comparable accuracy to single-phase baselines on standard cell-line benchmarks. Overall, these results demonstrate that learning structured and transferable representations from unlabeled molecular profiles can substantially reduce the amount of clinical supervision required for effective drug-response prediction, offering a practical pathway toward data-efficient preclinical-to-clinical translation. --- ### Polyglot-Lion: Efficient Multilingual ASR for Singapore via Balanced Fine-Tuning of Qwen3-ASR Authors: Quy-Anh Dang, Chris Ngo Published: 2026-03-17 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.16184v1 We present Polyglot-Lion, a family of compact multilingual automatic speech recognition (ASR) models tailored for the linguistic landscape of Singapore, covering English, Mandarin, Tamil, and Malay. Our models are obtained by fine-tuning Qwen3-ASR-0.6B and Qwen3-ASR-1.7B exclusively on publicly available speech corpora, using a balanced sampling strategy that equalizes the number of training utterances per language and deliberately omits language-tag conditioning so that the model learns to identify languages implicitly from audio. On 12 benchmarks spanning the four target languages, Polyglot-Lion-1.7B achieves an average error rate of 14.85, competitive with MERaLiON-2-10B-ASR (14.32) - a model 6x larger - while incurring a training cost of \$81 on a single RTX PRO 6000 GPU compared to \$18,862 for the 128-GPU baseline. Inference throughput is approximately 20x faster than MERaLiON at 0.10 s/sample versus 2.02 s/sample. These results demonstrate that linguistically balanced fine-tuning of moderate-scale pretrained models can yield deployment-ready multilingual ASR at a fraction of the cost of larger specialist systems. --- ### KidsNanny: A Two-Stage Multimodal Content Moderation Pipeline Integrating Visual Classification, Object Detection, OCR, and Contextual Reasoning for Child Safety Authors: Viraj Panchal, Tanmay Talsaniya, Parag Patel, Meet Patel Published: 2026-03-17 Categories: cs.CV, cs.CR arXiv: https://arxiv.org/abs/2603.16181v1 We present KidsNanny, a two-stage multimodal content moderation architecture for child safety. Stage 1 combines a vision transformer (ViT) with an object detector for visual screening (11.7 ms); outputs are routed as text not raw pixels to Stage 2, which applies OCR and a text based 7B language model for contextual reasoning (120 ms total pipeline). We evaluate on the UnsafeBench Sexual category (1,054 images) under two regimes: vision-only, isolating Stage 1, and multimodal, evaluating the full Stage 1+2 pipeline. Stage 1 achieves 80.27% accuracy and 85.39% F1 at 11.7 ms; vision-only baselines range from 59.01% to 77.04% accuracy. The full pipeline achieves 81.40% accuracy and 86.16% F1 at 120 ms, compared to ShieldGemma-2 (64.80% accuracy, 1,136 ms) and LlavaGuard (80.36% accuracy, 4,138 ms). To evaluate text-awareness, we filter two subsets: a text+visual subset (257 images) and a text-only subset (44 images where safety depends primarily on embedded text). On text-only images, KidsNanny achieves 100% recall (25/25 positives; small sample) and 75.76% precision; ShieldGemma-2 achieves 84% recall and 60% precision at 1,136 ms. Results suggest that dedicated OCR-based reasoning may offer recall-precision advantages on text-embedded threats at lower latency, though the small text-only subset limits generalizability. By documenting this architecture and evaluation methodology, we aim to contribute to the broader research effort on efficient multimodal content moderation for child safety. --- ### 360° Image Perception with MLLMs: A Comprehensive Benchmark and a Training-Free Method Authors: Huyen T. T. Tran, Van-Quang Nguyen, Farros Alferro, Kang-Jun Liu, Takayuki Okatani Published: 2026-03-17 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.16179v2 Multimodal Large Language Models (MLLMs) have shown impressive abilities in understanding and reasoning over conventional images. However, their perception of 360° images remains largely underexplored. Unlike conventional images, 360° images capture the entire surrounding environment, enabling holistic spatial reasoning but introducing challenges such as geometric distortion and complex spatial relations. To comprehensively assess MLLMs' capabilities to perceive 360° images, we introduce 360Bench, a Visual Question Answering (VQA) benchmark featuring 7K-resolution 360° images, seven representative (sub)tasks with annotations carefully curated by human annotators. Using 360Bench, we systematically evaluate seven MLLMs and six enhancement methods, revealing their shortcomings in 360° image perception. To address these challenges, we propose Free360, a training-free scene-graph-based framework for high-resolution 360° VQA. Free360 decomposes the reasoning process into modular steps, applies adaptive spherical image transformations to 360° images tailored to each step, and seamlessly integrates the resulting information into a unified graph representation for answer generation. Experiments show that Free360 consistently improves its base MLLM and provides a strong training-free solution for 360° VQA tasks. The source code and dataset will be publicly released upon acceptance. --- ### The Finetuner's Fallacy: When to Pretrain with Your Finetuning Data Authors: Christina Baek, Ricardo Pio Monti, David Schwab, Amro Abbas, Rishabh Adiga, Cody Blakeney, Maximilian Böther, Paul Burstein, Aldo Gael Carranza, Alvin Deng, Parth Doshi, Vineeth Dorna et al. Published: 2026-03-17 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.16177v2 Real-world model deployments demand strong performance on narrow domains where data is often scarce. Typically, practitioners finetune models to specialize them, but this risks overfitting to the domain and forgetting general knowledge. We study a simple strategy, specialized pretraining (SPT), where a small domain dataset, typically reserved for finetuning, is repeated starting from pretraining as a fraction of the total tokens. Across three specialized domains (ChemPile, MusicPile, and ProofPile), SPT improves domain performance and preserves general capabilities after finetuning compared to standard pretraining. In our experiments, SPT reduces the pretraining tokens needed to reach a given domain performance by up to 1.75x. These gains grow when the target domain is underrepresented in the pretraining corpus: on domains far from web text, a 1B SPT model outperforms a 3B standard pretrained model. Beyond these empirical gains, we derive overfitting scaling laws to guide practitioners in selecting the optimal domain-data repetition for a given pretraining compute budget. Our observations reveal the finetuner's fallacy: while finetuning may appear to be the cheapest path to domain adaptation, introducing specialized domain data during pretraining stretches its utility. SPT yields better specialized domain performance (via reduced overfitting across repeated exposures) and better general domain performance (via reduced forgetting during finetuning), ultimately achieving stronger results with fewer parameters and less total compute when amortized over inference. To get the most out of domain data, incorporate it as early in training as possible. --- ### MemX: A Local-First Long-Term Memory System for AI Assistants Authors: Lizheng Sun Published: 2026-03-17 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2603.16171v1 We present MemX, a local-first long-term memory system for AI assistants with stability-oriented retrieval design. MemX is implemented in Rust on top of libSQL and an OpenAI-compatible embedding API, providing persistent, searchable, and explainable memory for conversational agents. Its retrieval pipeline applies vector recall, keyword recall, Reciprocal Rank Fusion (RRF), four-factor re-ranking, and a low-confidence rejection rule that suppresses spurious recalls when no answer exists in the memory store. We evaluate MemX on two axes. First, two custom Chinese-language benchmark suites (43 queries, <=1,014 records) validate pipeline design: Hit@1=91.3% on a default scenario and 100% under high confusion, with conservative miss-query suppression. Second, the LongMemEval benchmark (500 queries, up to 220,349 records) quantifies system boundaries across four ability types and three storage granularities. At fact-level granularity the system reaches Hit@5=51.6% and MRR=0.380, doubling session-level performance, while temporal and multi-session reasoning remain challenging (<=43.6% Hit@5). FTS5 full-text indexing reduces keyword search latency by 1,100x at 100k-record scale, keeping end-to-end search under 90 ms. Unlike Mem0 and related work that targets end-to-end agent benchmarks, MemX focuses on a narrower, reproducible baseline: local-first deployment, structural simplicity, explainable retrieval, and stability-oriented design. --- ### Open-Source Reproduction and Explainability Analysis of Corrective Retrieval Augmented Generation Authors: Surya Vardhan Yalavarthi Published: 2026-03-17 Categories: cs.IR, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2603.16169v1 Corrective Retrieval Augmented Generation (CRAG) improves the robustness of RAG systems by evaluating retrieved document quality and triggering corrective actions. However, the original implementation relies on proprietary components including the Google Search API and closed model weights, limiting reproducibility. In this work, we present a fully open-source reproduction of CRAG, replacing proprietary web search with the Wikipedia API and the original LLaMA-2 generator with Phi-3-mini-4k-instruct. We evaluate on PopQA and ARC-Challenge, demonstrating that our open-source pipeline achieves comparable performance to the original system. Furthermore, we contribute the first explainability analysis of CRAG's T5-based retrieval evaluator using SHAP, revealing that the evaluator primarily relies on named entity alignment rather than semantic similarity. Our analysis identifies key failure modes including domain transfer limitations on science questions. All code and results are available at https://github.com/suryayalavarthi/crag-reproduction. --- ### SignNav: Leveraging Signage for Semantic Visual Navigation in Large-Scale Indoor Environments Authors: Jian Sun, Yuming Huang, He Li, Shuqi Xiao, Shenyan Guo, Maani Ghaffari, Qingbiao Li, Chengzhong Xu, Hui Kong Published: 2026-03-17 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2603.16166v1 Humans routinely leverage semantic hints provided by signage to navigate to destinations within novel Large-Scale Indoor (LSI) environments, such as hospitals and airport terminals. However, this capability remains underexplored within the field of embodied navigation. This paper introduces a novel embodied navigation task, SignNav, which requires the agent to interpret semantic hint from signage and reason about the subsequent action based on current observation. To facilitate research in this domain, we construct the LSI-Dataset for the training and evaluation of various SignNav agents. Dynamically changing semantic hints and sparse placement of signage in LSI environments present significant challenges to the SignNav task. To address these challenges, we propose the Spatial-Temporal Aware Transformer (START) model for end-to-end decision-making. The spatial-aware module grounds the semantic hint of signage into physical world, while the temporal-aware module captures long-range dependencies between historical states and current observation. Leveraging a two-stage training strategy with Dataset Aggregation (DAgger), our approach achieves state-of-the-art performance, recording an 80% Success Rate (SR) and 0.74 NDTW on val-unseen split. Real-world deployment further demonstrates the practicality of our method in physical environment without pre-built map. --- ### Homogeneous and Heterogeneous Consistency progressive Re-ranking for Visible-Infrared Person Re-identification Authors: Yiming Wang Published: 2026-03-17 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.16165v1 Visible-infrared person re-identification faces greater challenges than traditional person re-identification due to the significant differences between modalities. In particular, the differences between these modalities make effective matching even more challenging, mainly because existing re-ranking algorithms cannot simultaneously address the intra-modal variations and inter-modal discrepancy in cross-modal person re-identification. To address this problem, we propose a novel Progressive Modal Relationship Re-ranking method consisting of two modules, called heterogeneous and homogeneous consistency re-ranking(HHCR). The first module, heterogeneous consistency re-ranking, explores the relationship between the query and the gallery modalities in the test set. The second module, homogeneous consistency reranking, investigates the intrinsic relationship within each modality between the query and the gallery in the test set. Based on this, we propose a baseline for cross-modal person re-identification, called a consistency re-ranking inference network (CRI). We conducted comprehensive experiments demonstrating that our proposed re-ranking method is generalized, and both the re-ranking and the baseline achieve state-of-the-art performance. --- ### STARK: Spatio-Temporal Attention for Representation of Keypoints for Continuous Sign Language Recognition Authors: Suvajit Patra, Soumitra Samanta Published: 2026-03-17 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2603.16163v1 Continuous Sign Language Recognition (CSLR) is a crucial task for understanding the languages of deaf communities. Contemporary keypoint-based approaches typically rely on spatio-temporal encoding, where spatial interactions among keypoints are modeled using Graph Convolutional Networks or attention mechanisms, while temporal dynamics are captured using 1D convolutional networks. However, such designs often introduce a large number of parameters in both the encoder and the decoder. This paper introduces a unified spatio-temporal attention network that computes attention scores both spatially (across keypoints) and temporally (within local windows), and aggregates features to produce a local context-aware spatio-temporal representation. The proposed encoder contains approximately $70-80\%$ fewer parameters than existing state-of-the-art models while achieving comparable performance to keypoint-based methods on the Phoenix-14T dataset. --- ### SQL-ASTRA: Alleviating Sparse Feedback in Agentic SQL via Column-Set Matching and Trajectory Aggregation Authors: Long Li, Zhijian Zhou, Jiangxuan Long, Peiyang Liu, Weidi Xu, Zhe Wang, Shirui Pan, Chao Qu Published: 2026-03-17 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.16161v1 Agentic Reinforcement Learning (RL) shows promise for complex tasks, but Text-to-SQL remains mostly restricted to single-turn paradigms. A primary bottleneck is the credit assignment problem. In traditional paradigms, rewards are determined solely by the final-turn feedback, which ignores the intermediate process and leads to ambiguous credit evaluation. To address this, we propose Agentic SQL, a framework featuring a universal two-tiered reward mechanism designed to provide effective trajectory-level evaluation and dense step-level signals. First, we introduce Aggregated Trajectory Reward (ATR) to resolve multi-turn credit assignment. Using an asymmetric transition matrix, ATR aggregates process-oriented scores to incentivize continuous improvement. Leveraging Lyapunov stability theory, we prove ATR acts as an energy dissipation operator, guaranteeing a cycle-free policy and monotonic convergence. Second, Column-Set Matching Reward (CSMR) provides immediate step-level rewards to mitigate sparsity. By executing queries at each turn, CSMR converts binary (0/1) feedback into dense (0, 1) signals based on partial correctness. Evaluations on BIRD show a 5% gain over binary-reward GRPO. Notably, our approach outperforms SOTA Arctic-Text2SQL-R1-7B on BIRD and Spider 2.0 using identical models, propelling Text-to-SQL toward a robust multi-turn agent paradigm. --- ### Segmentation-before-Staining Improves Structural Fidelity in Virtual IHC-to-Multiplex IF Translation Authors: Junhyeok Lee, Han Jang, Heeseong Eum, Joon Jang, Kyu Sung Choi Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16160v1 Multiplex immunofluorescence (mIF) enables simultaneous single-cell quantification of multiple biomarkers within intact tissue architecture, yet its high reagent cost, multi-round staining protocols, and need for specialized imaging platforms limit routine clinical adoption. Virtual staining can synthesize mIF channels from widely available brightfield immunohistochemistry (IHC), but current translators optimize pixel-level fidelity without explicitly constraining nuclear morphology. In pathology, this gap is clinically consequential: subtle distortions in nuclei count, shape, or spatial arrangement propagate directly to quantification endpoints such as the Ki67 proliferation index, where errors of a few percent can shift treatment-relevant risk categories. This work introduces a supervision-free, architecture-agnostic conditioning strategy that injects a continuous cell probability map from a pretrained nuclei segmentation foundation model as an explicit input prior, together with a variance-preserving regularization term that matches local intensity statistics to maintain cell-level heterogeneity in synthesized fluorescence channels. The soft prior retains gradient-level boundary information lost by binary thresholding, providing a richer conditioning signal without task-specific tuning. Controlled experiments across Pix2Pix with U-Net and ResNet generators, deterministic regression U-Net, and conditional diffusion on two independent datasets demonstrate consistent improvements in nuclei count fidelity and perceptual quality, as the sole modifications. Code will be made publicly available upon acceptance. --- ### AI-Generated Figures in Academic Publishing: Policies, Tools, and Practical Guidelines Authors: Davie Chen Published: 2026-03-17 Categories: cs.CV, cs.CY arXiv: https://arxiv.org/abs/2603.16159v1 The rapid advancement of generative AI has introduced a new class of tools capable of producing publication-quality scientific figures, graphical abstracts, and data visualizations. However, academic publishers have responded with inconsistent and often ambiguous policies regarding AI-generated imagery. This paper surveys the current stance of major journals and publishers -- including Nature, Science, Cell Press, Elsevier, and PLOS -- on the use of AI-generated figures. We identify key concerns raised by publishers, including reproducibility, authorship attribution, and potential for visual misinformation. Drawing on practical examples from tools such as SciDraw, an AI-powered platform designed specifically for scientific illustration, we propose a set of best-practice guidelines for researchers seeking to use AI figure-generation tools in a compliant and transparent manner. Our findings suggest that, with appropriate disclosure and quality control, AI-generated figures can meaningfully accelerate scientific communication without compromising integrity. --- ### Execution-Grounded Credit Assignment for GRPO in Code Generation Authors: Abhijit Kumar, Natalya Kumar, Shikhar Gupta Published: 2026-03-17 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.16158v1 Critic-free reinforcement learning with verifiable rewards (RLVR) improves code generation by optimizing unit-test pass rates, but GRPO-style updates suffer from coarse credit assignment: a single outcome signal is spread uniformly across long programs even when failure stems from a localized semantic error. We propose Execution-Grounded Credit Assignment (EGCA), which localizes GRPO updates using execution traces. For programs that satisfy algorithmic constraints but fail tests, EGCA executes the candidate and a canonical reference solution (curated once offline; used for analysis, not supervision) under identical instrumentation, identifies the earliest semantic divergence, and assigns advantage only to the corresponding token span while masking downstream tokens. EGCA is a drop-in modification requiring no critic, auxiliary loss, or learned verifier, yielding 82.1% pass@1 on HumanEval (+3.1 over GRPO) and 68.9% on MBPP (+1.5) with 18% wall-clock overhead. --- ### DyJR: Preserving Diversity in Reinforcement Learning with Verifiable Rewards via Dynamic Jensen-Shannon Replay Authors: Long Li, Zhijian Zhou, Tianyi Wang, Weidi Xu, Zuming Huang, Wei Chu, Zhe Wang, Shirui Pan, Chao Qu, Yuan Qi Published: 2026-03-17 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.16157v1 While Reinforcement Learning (RL) enhances Large Language Model reasoning, on-policy algorithms like GRPO are sample-inefficient as they discard past rollouts. Existing experience replay methods address this by reusing accurate samples for direct policy updates, but this often incurs high computational costs and causes mode collapse via overfitting. We argue that historical data should prioritize sustaining diversity rather than simply reinforcing accuracy. To this end, we propose Dynamic Jensen-Shannon Replay (DyJR), a simple yet effective regularization framework using a dynamic reference distribution from recent trajectories. DyJR introduces two innovations: (1) A Time-Sensitive Dynamic Buffer that uses FIFO and adaptive sizing to retain only temporally proximal samples, synchronizing with model evolution; and (2) Jensen-Shannon Divergence Regularization, which replaces direct gradient updates with a distributional constraint to prevent diversity collapse. Experiments on mathematical reasoning and Text-to-SQL benchmarks demonstrate that DyJR significantly outperforms GRPO as well as baselines such as RLEP and Ex-GRPO, while maintaining training efficiency comparable to the original GRPO. Furthermore, from the perspective of Rank-$k$ token probability evolution, we show that DyJR enhances diversity and mitigates over-reliance on Rank-1 tokens, elucidating how specific sub-modules of DyJR influence the training dynamics. --- ### GATS: Gaussian Aware Temporal Scaling Transformer for Invariant 4D Spatio-Temporal Point Cloud Representation Authors: Jiayi Tian, Jiaze Wang Published: 2026-03-17 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.16154v1 Understanding 4D point cloud videos is essential for enabling intelligent agents to perceive dynamic environments. However, temporal scale bias across varying frame rates and distributional uncertainty in irregular point clouds make it highly challenging to design a unified and robust 4D backbone. Existing CNN or Transformer based methods are constrained either by limited receptive fields or by quadratic computational complexity, while neglecting these implicit distortions. To address this problem, we propose a novel dual invariant framework, termed \textbf{Gaussian Aware Temporal Scaling (GATS)}, which explicitly resolves both distributional inconsistencies and temporal. The proposed \emph{Uncertainty Guided Gaussian Convolution (UGGC)} incorporates local Gaussian statistics and uncertainty aware gating into point convolution, thereby achieving robust neighborhood aggregation under density variation, noise, and occlusion. In parallel, the \emph{Temporal Scaling Attention (TSA)} introduces a learnable scaling factor to normalize temporal distances, ensuring frame partition invariance and consistent velocity estimation across different frame rates. These two modules are complementary: temporal scaling normalizes time intervals prior to Gaussian estimation, while Gaussian modeling enhances robustness to irregular distributions. Our experiments on mainstream benchmarks MSR-Action3D (\textbf{+6.62\%} accuracy), NTU RGBD (\textbf{+1.4\%} accuracy), and Synthia4D (\textbf{+1.8\%} mIoU) demonstrate significant performance gains, offering a more efficient and principled paradigm for invariant 4D point cloud video understanding with superior accuracy, robustness, and scalability compared to Transformer based counterparts. --- ### HIPO: Instruction Hierarchy via Constrained Reinforcement Learning Authors: Keru Chen, Jun Luo, Sen Lin, Yingbin Liang, Alvaro Velasquez, Nathaniel Bastian, Shaofeng Zou Published: 2026-03-17 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2603.16152v1 Hierarchical Instruction Following (HIF) refers to the problem of prompting large language models with a priority-ordered stack of instructions. Standard methods like RLHF and DPO typically fail in this problem since they mainly optimize for a single objective, failing to explicitly enforce system prompt compliance. Meanwhile, supervised fine-tuning relies on mimicking filtered, compliant data, which fails to establish the priority asymmetry at the algorithmic level. In this paper, we introduce \textsc{HIPO}, a novel alignment framework that formulates HIF as a Constrained Markov Decision Process. \textsc{HIPO} elevates system prompts from mere input context to strict algorithmic boundaries. Using a primal-dual safe reinforcement learning approach, the algorithm dynamically enforces system prompt compliance as an explicit constraint, maximizing user utility strictly within this feasible region. Extensive evaluations across diverse model architectures (e.g., Qwen, Phi, Llama) demonstrate that \textsc{HIPO} significantly improves both system compliance and user utility. Furthermore, mechanistic analysis reveals that this constrained optimization autonomously drives the model to shift its attention toward long-range system tokens, providing a principled foundation for reliable LLM deployment in complex workflows. --- ### EFF-Grasp: Energy-Field Flow Matching for Physics-Aware Dexterous Grasp Generation Authors: Yukun Zhao, Zichen Zhong, Yongshun Gong, Yilong Yin, Haoliang Sun Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16151v1 Denoising generative models have recently become the dominant paradigm for dexterous grasp generation, owing to their ability to model complex grasp distributions from large-scale data. However, existing diffusion-based methods typically formulate generation as a stochastic differential equation (SDE), which often requires many sequential denoising steps and introduces trajectory instability that can lead to physically infeasible grasps. In this paper, we propose EFF-Grasp, a novel Flow-Matching-based framework for physics-aware dexterous grasp generation. Specifically, we reformulate grasp synthesis as a deterministic ordinary differential equation (ODE) process, which enables efficient and stable generation through smooth probability flows. To further enforce physical feasibility, we introduce a training-free physics-aware energy guidance strategy. Our method defines an energy-guided target distribution using adapted explicit physical energy functions that capture key grasp constraints, and estimates the corresponding guidance term via a local Monte Carlo approximation during inference. In this way, EFF-Grasp dynamically steers the generation trajectory toward physically feasible regions without requiring additional physics-based training or simulation feedback. Extensive experiments on five benchmark datasets show that EFF-Grasp achieves superior performance in grasp quality and physical feasibility, while requiring substantially fewer sampling steps than diffusion-based baselines. --- ### NeuronSpark: A Spiking Neural Network Language Model with Selective State Space Dynamics Authors: Zhengzheng Tang Published: 2026-03-17 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.16148v1 We ask whether a pure spiking backbone can learn large-scale language modeling from random initialization, without Transformer distillation. We introduce NeuronSpark, a 0.9B-parameter SNN language model trained with next-token prediction and surrogate gradients. The model combines selective state-space spiking dynamics, leakage-current inter-layer communication, PonderNet adaptive timesteps, fused Triton PLIF kernels, and stabilization techniques (residual centering, lateral-inhibition normalization, and natural-gradient compensation). Under a constrained budget (about 1.4B pretraining tokens and 6.5K SFT steps), NeuronSpark-0.9B reaches 3.6 pretraining loss and shows early multi-turn dialogue behavior after SFT. These results support the feasibility of end-to-end language modeling with a pure SNN architecture at this scale. --- ### Deep Adaptive Model-Based Design of Experiments Authors: Arno Strouwen, Sebastian Micluţa-Câmpeanu Published: 2026-03-17 Categories: stat.ML, cs.LG, eess.SY, stat.ME arXiv: https://arxiv.org/abs/2603.16146v2 Model-based design of experiments (MBDOE) is essential for efficient parameter estimation in nonlinear dynamical systems. However, conventional adaptive MBDOE requires costly posterior inference and design optimization between each experimental step, precluding real-time applications. We address this by combining Deep Adaptive Design (DAD), which amortizes sequential design into a neural network policy trained offline, with differentiable mechanistic models. For dynamical systems with known governing equations but uncertain parameters, we extend sequential contrastive training objectives to handle nuisance parameters and propose a transformer-based policy architecture that respects the temporal structure of dynamical systems. We demonstrate the approach on four systems of increasing complexity: a fed-batch bioreactor with Monod kinetics, a Haldane bioreactor with uncertain substrate inhibition, a two-compartment pharmacokinetic model with nuisance clearance parameters, and a DC motor for real-time deployment. --- ### Structure-Aware Multimodal LLM Framework for Trustworthy Near-Field Beam Prediction Authors: Mengyuan Li, Qianfan Lu, Jiachen Tian, Hongjun Hu, Yu Han, Xiao Li, Chao-kai Wen, Shi Jin Published: 2026-03-17 Categories: eess.SP, cs.AI arXiv: https://arxiv.org/abs/2603.16143v1 In near-field extremely large-scale multiple-input multiple-output (XL-MIMO) systems, spherical wavefront propagation expands the traditional beam codebook into the joint angular-distance domain, rendering conventional beam training prohibitively inefficient, especially in complex 3-dimensional (3D) low-altitude environments. Furthermore, since near-field beam variations are deeply coupled not only with user positions but also with the physical surroundings, precise beam alignment demands profound environmental understanding capabilities. To address this, we propose a large language model (LLM)-driven multimodal framework that fuses historical GPS data, RGB image, LiDAR data, and strategically designed task-specific textual prompts. By utilizing the powerful emergent reasoning and generalization capabilities of the LLM, our approach learns complex spatial dynamics to achieve superior environmental comprehension... --- ### Parametric Social Identity Injection and Diversification in Public Opinion Simulation Authors: Hexi Wang, Yujia Zhou, Bangde Du, Qingyao Ai, Yiqun Liu Published: 2026-03-17 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.16142v2 Large language models (LLMs) have recently been adopted as synthetic agents for public opinion simulation, offering a promising alternative to costly and slow human surveys. Despite their scalability, current LLM-based simulation methods fail to capture social diversity, producing flattened inter-group differences and overly homogeneous responses across demographic groups. We identify this limitation as a Diversity Collapse phenomenon in LLM hidden representations, where distinct social identities become increasingly indistinguishable across layers. Motivated by this observation, we propose Parametric Social Identity Injection (PSII), a general framework that injects explicit, parametric representations of demographic attributes and value orientations directly into intermediate hidden states of LLMs. Unlike prompt-based persona conditioning, PSII enables fine-grained and controllable identity modulation at the representation level. Extensive experiments on the World Values Survey using multiple open-source LLMs show that PSII significantly improves distributional fidelity and diversity, reducing KL divergence to real-world survey data while enhancing overall diversity. This work provides new insights into representation-level control of LLM agents and advances scalable, diversity-aware public opinion simulation. --- ### Communication-Aware Multi-Agent Reinforcement Learning for Decentralized Cooperative UAV Deployment Authors: Enguang Fan, Yifan Chen, Zihan Shan, Klara Nahrstedt, Matthew Caesar, Jae Kim Published: 2026-03-17 Categories: cs.MA, cs.LG, cs.NI arXiv: https://arxiv.org/abs/2603.16141v2 Autonomous Unmanned Aerial Vehicle (UAV) swarms are increasingly used as rapidly deployable aerial relays and sensing platforms, yet practical deployments must operate under partial observability and intermittent peer-to-peer connectivity. We present a graph-based multi-agent reinforcement learning framework trained under centralized training with decentralized execution (CTDE): a centralized critic and global state are available only during training, while each UAV executes a shared policy using local observations and messages from nearby neighbors. Under restricted communication, neighbor relations are induced by an SNR-threshold connectivity graph. Our architecture encodes local agent state and nearby entities with an agent-entity attention module and aggregates inter-UAV messages with neighbor self-attention over a signal-quality-limited communication graph defined by a channel model. We evaluate the framework on a cooperative relay-deployment task, DroneConnect. Experimental results show that the proposed method achieves an approximately 12% increase in target coverage over MAPPO under restricted communication and partial observability, while remaining competitive with a mixed-integer linear programming (MILP)-based offline upper bound with full node observability. --- ### Noisy Data is Destructive to Reinforcement Learning with Verifiable Rewards Authors: Yuxuan Zhu, Daniel Kang Published: 2026-03-17 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.16140v2 Reinforcement learning with verifiable rewards (RLVR) has driven recent capability advances of large language models across various domains. Recent studies suggest that improved RLVR algorithms allow models to learn effectively from incorrect annotations, achieving performance comparable to learning from clean data. In this work, we show that these findings are invalid because the claimed 100% noisy training data is "contaminated" with clean data. After rectifying the dataset with a rigorous re-verification pipeline, we demonstrate that noise is destructive to RLVR. We show that existing RLVR algorithm improvements fail to mitigate the impact of noise, achieving similar performance to that of the basic GRPO. Furthermore, we find that the model trained on truly incorrect annotations performs 8-10% worse than the model trained on clean data across mathematical reasoning benchmarks. Finally, we show that these findings hold for real-world noise in Text2SQL tasks, where training on real-world, human annotation errors cause 5-12% lower accuracy than clean data. Our results show that current RLVR methods cannot yet compensate for poor data quality. High-quality data remains essential. --- ### Rethinking UMM Visual Generation: Masked Modeling for Efficient Image-Only Pre-training Authors: Peng Sun, Jun Xie, Tao Lin Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16139v1 Unified Multimodal Models (UMMs) are often constrained by the pre-training of their $\textbf{visual generation components}$, which typically relies on inefficient paradigms and scarce, high-quality text-image paired data. In this paper, we systematically analyze pre-training recipes for $\textbf{UMM visual generation}$ and identify these two issues as the major bottlenecks. To address them, we propose $\textbf{Image-Only Training for UMMs (IOMM)}$, a data-efficient two-stage training framework. The first stage pre-trains the visual generative component $\textbf{exclusively}$ using abundant unlabeled image-only data, thereby removing the dependency on paired data $\textbf{for this costly phase}$. The second stage fine-tunes the model using a mixture of unlabeled images and a small curated set of text-image pairs, leading to improved instruction alignment and generative quality. Extensive experiments show that IOMM not only improves training efficiency but also achieves state-of-the-art (SOTA) performance. For example, our IOMM-B (3.6B) model was trained from scratch using only $\sim \textbf{1050}$ H800 GPU hours (with the vast majority, $\textbf{1000}$ hours, dedicated to the efficient $\textbf{image-only pre-training stage}$). It achieves $\textbf{0.89}$ on GenEval and $\textbf{0.55}$ on WISE--surpassing strong baselines such as BAGEL-7B (0.82 & 0.55) and BLIP3-o-4B (0.84 & 0.50). Code is available $\href{https://github.com/LINs-lab/IOMM}{https://github.com/LINs-lab/IOMM}$. --- ### Answer Bubbles: Information Exposure in AI-Mediated Search Authors: Michelle Huang, Agam Goyal, Koustuv Saha, Eshwar Chandrasekharan Published: 2026-03-17 Categories: cs.IR, cs.CL arXiv: https://arxiv.org/abs/2603.16138v2 Generative search systems are increasingly replacing link-based retrieval with AI-generated summaries, yet little is known about how these systems differ in sources, language, and fidelity to cited material. We examine responses to 11,000 real search queries across five systems---vanilla GPT, Search GPT, Perplexity Search with Grok, Google AI Overviews, and traditional Google Search---at three levels: source diversity, linguistic characterization of the generated summary, and source-summary fidelity. We find that generative search systems exhibit significant \textit{source-selection} biases in their citations, favoring certain sources over others. Incorporating search also selectively attenuates epistemic markers, reducing hedging by up to 60\% while preserving confidence language in the AI-generated summaries. At the same time, AI summaries further compound the citation biases: Wikipedia and longer sources are disproportionately overrepresented, whereas cited social media content and negatively framed sources are substantially underrepresented. Our findings highlight the potential for \textit{answer bubbles}, in which identical queries yield structurally different information realities across systems, with implications for user trust, source visibility, and the transparency of AI-mediated information access. --- ### Answer Bubbles: Information Exposure in AI-Mediated Search Authors: Michelle Huang, Agam Goyal, Koustuv Saha, Eshwar Chandrasekharan Published: 2026-03-17 Categories: cs.IR, cs.CL arXiv: https://arxiv.org/abs/2603.16138v1 Generative search systems are increasingly replacing link-based retrieval with AI-generated summaries, yet little is known about how these systems differ in sources, language, and fidelity to cited material. We examine responses to 11,000 real search queries across four systems -- vanilla GPT, Search GPT, Google AI Overviews, and traditional Google Search -- at three levels: source diversity, linguistic characterization of the generated summary, and source-summary fidelity. We find that generative search systems exhibit significant \textit{source-selection} biases in their citations, favoring certain sources over others. Incorporating search also selectively attenuates epistemic markers, reducing hedging by up to 60\% while preserving confidence language in the AI-generated summaries. At the same time, AI summaries further compound the citation biases: Wikipedia and longer sources are disproportionately overrepresented, whereas cited social media content and negatively framed sources are substantially underrepresented. Our findings highlight the potential for \textit{answer bubbles}, in which identical queries yield structurally different information realities across systems, with implications for user trust, source visibility, and the transparency of AI-mediated information access. --- ### SIA: A Synthesize-Inject-Align Framework for Knowledge-Grounded and Secure E-commerce Search LLMs with Industrial Deployment Authors: Zhouwei Zhai, Mengxiang Chen, Anmeng Zhang Published: 2026-03-17 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.16137v2 Large language models offer transformative potential for e-commerce search by enabling intent-aware recommendations. However, their industrial deployment is hindered by two critical challenges: (1) knowledge hallucination due to insufficient encoding of dynamic, fine-grained product knowledge, and (2) security vulnerabilities under jailbreak attacks that threaten compliance. To address these issues, we propose SIA--a Synthesize-Inject-Align framework for building knowledgeable and secure e-commerce search LLMs. Our approach first synthesizes high-quality natural language corpus by combining structured knowledge graphs with unstructured behavioral logs, augmented with reasoning chains and safety-aware data. We then introduce a parameter-efficient pre-training strategy based on Depth Up-Scaling to inject domain knowledge while preserving general capabilities. Finally, a dual-path alignment method via multi-task instruction tuning and adversarial training strengthens both task performance and safety robustness. The framework has been deployed at JD.com, China's largest self-operated e-commerce platform, where A/B tests across five core search scenarios demonstrate significant improvements in key business metrics, validating its industrial effectiveness and scalability. --- ### When Generative Augmentation Hurts: A Benchmark Study of GAN and Diffusion Models for Bias Correction in AI Classification Systems Authors: Shesh Narayan Gupta, Nik Bear Brown Published: 2026-03-17 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.16134v1 Generative models are widely used to compensate for class imbalance in AI training pipelines, yet their failure modes under low-data conditions are poorly understood. This paper reports a controlled benchmark comparing three augmentation strategies applied to a fine-grained animal classification task: traditional transforms, FastGAN, and Stable Diffusion 1.5 fine-tuned with Low-Rank Adaptation (LoRA). Using the Oxford-IIIT Pet Dataset with eight artificially underrepresented breeds, we find that FastGAN augmentation does not merely underperform at very low training set sizes but actively increases classifier bias, with a statistically significant large effect across three random seeds (bias gap increase: +20.7%, Cohen's d = +5.03, p = 0.013). The effect size here is large enough to give confidence in the direction of the finding despite the small number of seeds. Feature embedding analysis using t-distributed Stochastic Neighbor Embedding reveals that FastGAN images for severe-minority breeds form tight isolated clusters outside the real image distribution, a pattern consistent with mode collapse. Stable Diffusion with Low-Rank Adaptation produced the best results overall, achieving the highest macro F1 (0.9125 plus or minus 0.0047) and a 13.1% reduction in the bias gap relative to the unaugmented baseline. The data suggest a sample-size boundary somewhere between 20 and 50 training images per class below which GAN augmentation becomes harmful in this setting, though further work across additional domains is needed to establish where that boundary sits more precisely. All experiments run on a consumer-grade GPU with 6 to 8 GB of memory, with no cloud compute required. --- ### DualPrim: Compact 3D Reconstruction with Positive and Negative Primitives Authors: Xiaoxu Meng, Zhongmin Chen, Bo Yang, Weikai Chen, Weixiao Liu, Lin Gao Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16133v1 Neural reconstructions often trade structure for fidelity, yielding dense and unstructured meshes with irregular topology and weak part boundaries that hinder editing, animation, and downstream asset reuse. We present DualPrim, a compact and structured 3D reconstruction framework. Unlike additive-only implicit or primitive methods, DualPrim represents shapes with positive and negative superquadrics: the former builds the bases while the latter carves local volumes through a differentiable operator, enabling topology-aware modeling of holes and concavities. This additive-subtractive design increases the representational power without sacrificing compactness or differentiability. We embed DualPrim in a volumetric differentiable renderer, enabling end-to-end learning from multi-view images and seamless mesh export via closed-form boolean difference. Empirically, DualPrim delivers state-of-the-art accuracy and produces compact, structured, and interpretable outputs that better satisfy downstream needs than additive-only alternatives. --- ### SciZoom: A Large-scale Benchmark for Hierarchical Scientific Summarization across the LLM Era Authors: Han Jang, Junhyeok Lee, Kyu Sung Choi Published: 2026-03-17 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.16131v2 The explosive growth of AI research has created unprecedented information overload, increasing the demand for scientific summarization at multiple levels of granularity beyond traditional abstracts. While LLMs are increasingly adopted for summarization, existing benchmarks remain limited in scale, target only a single granularity, and predate the LLM era. Moreover, since the release of ChatGPT in November 2022, researchers have rapidly adopted LLMs for drafting manuscripts themselves, fundamentally transforming scientific writing, yet no resource exists to analyze how this writing has evolved. To bridge these gaps, we introduce SciZoom, a benchmark comprising 44,946 papers from four top-tier ML venues (NeurIPS, ICLR, ICML, EMNLP) spanning 2020 to 2025, explicitly stratified into Pre-LLM and Post-LLM eras. SciZoom provides three hierarchical summarization targets (Abstract, Contributions, and TL;DR) achieving compression ratios up to 600:1, enabling both multi-granularity summarization research and temporal mining of scientific writing patterns. Our linguistic analysis reveals striking shifts in phrase patterns (up to 10x for formulaic expressions) and rhetorical style (23% decline in hedging), suggesting that LLM-assisted writing produces more confident yet homogenized prose. SciZoom serves as both a challenging benchmark and a unique resource for mining the evolution of scientific discourse in the generative AI era. Our code and dataset are publicly available on GitHub (https://github.com/janghana/SciZoom) and Hugging Face (https://huggingface.co/datasets/hanjang/SciZoom), respectively. --- ### EPOFusion: Exposure aware Progressive Optimization Method for Infrared and Visible Image Fusion Authors: Zhiwei Wang, Defeng He, Li Zhao, Xiaoqin Zhang, Yuxing Li, Edmund Y. Lam Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16130v5 Overexposure caused by strong daylight and oncoming headlights frequently overwhelms visible sensors, resulting in critical information loss in visual perception. Infrared and visible image fusion can compensate for such degradation via multimodal complementarity. However, most fusion methods lack region-aware optimization for overexposed areas and cannot effectively exploit infrared cues in saturated regions, resulting in insufficient infrared detail preservation or redundant information in the fused results. To address this, we propose EPOFusion, an exposure-aware fusion framework. It employs a spatial guidance module to identify regions requiring infrared compensation, together with a region-aware fusion loss to strengthen informative infrared structures. In addition, an iterative feature refinement head equipped with a multiscale context fusion module progressively refines fused representations, enabling effective integration of complementary infrared information while maintaining visual consistency in normally exposed regions. The infrared and visible overexposure (IVOE) dataset consists of a synthetic training subset providing infrared-compensation supervision and a real-world subset for fusion and downstream perception evaluation under authentic overexposure. EPOFusion demonstrates superior VIF and $Q^{AB/F}$ performance with favorable visual quality, improving $Q^{AB/F}$ by 10.7% over the existing overexposure-oriented fusion baseline, while further improving downstream mIoU and mAP50 by 5.6% and 6.5%, respectively. Code, results, and the IVOE dataset will be made available at https://warren-wzw.github.io/EPOFusion/. --- ### EPOFusion: Exposure aware Progressive Optimization Method for Infrared and Visible Image Fusion Authors: Zhiwei Wang, Defeng He, Li Zhao, Xiaoqin Zhang, Yuxing Li, Edmund Y. Lam Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16130v4 Overexposure caused by strong daylight and oncoming headlights frequently overwhelms visible sensors, resulting in critical information loss in visual perception. Infrared and visible image fusion can compensate for such degradation via multimodal complementarity. However, most fusion methods lack region-aware optimization for overexposed areas and cannot effectively exploit infrared cues in saturated regions, resulting in insufficient infrared detail preservation or redundant information in the fused results. To address this, we propose EPOFusion, an exposure-aware fusion framework. It uses a spatial guidance module to selectively preserve informative infrared cues in overexposed regions. In addition, an iterative decoding head equipped with a multiscale context fusion module progressively refines fused representations, enabling effective infrared compensation in degraded regions while maintaining visual consistency in normal regions. The infrared and visible overexposure (IVOE) dataset is constructed with a synthetic training subset for controlled supervision and a real-world test subset for generalization assessment, supporting exposure-aware learning and evaluation. Extensive experiments on MSRS, FMB, and the proposed IVOE benchmark show that EPOFusion improves information preservation and visual fidelity, achieving an average full-image MI gain of 28.7% over the best competing methods. Qualitative results further demonstrate effective compensation in saturated regions, and downstream evaluations confirm its benefits under challenging overexposed conditions. Code, results, and the IVOE dataset will be made available at https://github.com/warren-wzw/EPOFusion. --- ### Boosting Quantitive and Spatial Awareness for Zero-Shot Object Counting Authors: Da Zhang, Bingyu Li, Feiyu Wang, Zhiyuan Zhao, Junyu Gao Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16129v1 Zero-shot object counting (ZSOC) aims to enumerate objects of arbitrary categories specified by text descriptions without requiring visual exemplars. However, existing methods often treat counting as a coarse retrieval task, suffering from a lack of fine-grained quantity awareness. Furthermore, they frequently exhibit spatial insensitivity and degraded generalization due to feature space distortion during model adaptation.To address these challenges, we present \textbf{QICA}, a novel framework that synergizes \underline{q}uantity percept\underline{i}on with robust spatial \underline{c}ast \underline{a}ggregation. Specifically, we introduce a Synergistic Prompting Strategy (\textbf{SPS}) that adapts vision and language encoders through numerically conditioned prompts, bridging the gap between semantic recognition and quantitative reasoning. To mitigate feature distortion, we propose a Cost Aggregation Decoder (\textbf{CAD}) that operates directly on vision-text similarity maps. By refining these maps through spatial aggregation, CAD prevents overfitting while preserving zero-shot transferability. Additionally, a multi-level quantity alignment loss ($\mathcal{L}_{MQA}$) is employed to enforce numerical consistency across the entire pipeline. Extensive experiments on FSC-147 demonstrate competitive performance, while zero-shot evaluation on CARPK and ShanghaiTech-A validates superior generalization to unseen domains. --- ### Social Simulacra in the Wild: AI Agent Communities on Moltbook Authors: Agam Goyal, Olivia Pal, Hari Sundaram, Eshwar Chandrasekharan, Koustuv Saha Published: 2026-03-17 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.16128v3 As autonomous LLM-based agents increasingly populate social platforms, understanding the dynamics of AI-agent communities becomes essential for both communication research and platform governance. We present the first large-scale empirical comparison of AI-agent and human online communities, analyzing 73,899 Moltbook and 189,838 Reddit posts across five matched communities. Structurally, we find that Moltbook exhibits extreme participation inequality (Gini = 0.84 vs. 0.47) and high cross-community author overlap (33.8% vs. 0.5%). In terms of linguistic attributes, content generated by AI-agents is emotionally flattened, cognitively shifted toward assertion over exploration, and socially detached. These differences give rise to apparent community-level homogenization, but we show this is primarily a structural artifact of shared authorship. At the author level, individual agents are more identifiable than human users, driven by outlier stylistic profiles amplified by their extreme posting volume. As AI-mediated communication reshapes online discourse, our work offers an empirical foundation for understanding how multi-agent interaction gives rise to collective communication dynamics distinct from those of human communities. --- ### Social Simulacra in the Wild: AI Agent Communities on Moltbook Authors: Agam Goyal, Olivia Pal, Hari Sundaram, Eshwar Chandrasekharan, Koustuv Saha Published: 2026-03-17 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.16128v2 As autonomous LLM-based agents increasingly populate social platforms, understanding the dynamics of AI-agent communities becomes essential for both communication research and platform governance. We present the first large-scale empirical comparison of AI-agent and human online communities, analyzing 73,899 Moltbook and 189,838 Reddit posts across five matched communities. Structurally, we find that Moltbook exhibits extreme participation inequality (Gini = 0.84 vs. 0.47) and high cross-community author overlap (33.8\% vs. 0.5\%). In terms of linguistic attributes, content generated by AI-agents is emotionally flattened, cognitively shifted toward assertion over exploration, and socially detached. These differences give rise to apparent community-level homogenization, but we show this is primarily a structural artifact of shared authorship. At the author level, individual agents are more identifiable than human users, driven by outlier stylistic profiles amplified by their extreme posting volume. As AI-mediated communication reshapes online discourse, our work offers an empirical foundation for understanding how multi-agent interaction gives rise to collective communication dynamics distinct from those of human communities. --- ### Pre-training LLM without Learning Rate Decay Enhances Supervised Fine-Tuning Authors: Kazuki Yano, Shun Kiyono, Sosuke Kobayashi, Sho Takase, Jun Suzuki Published: 2026-03-17 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2603.16127v1 We investigate the role of learning rate scheduling in the large-scale pre-training of large language models, focusing on its influence on downstream performance after supervised fine-tuning (SFT). Decay-based learning rate schedulers are widely used to minimize pre-training loss. However, despite their widespread use, how these schedulers affect performance after SFT remains underexplored. In this paper, we examine Warmup-Stable-Only (WSO), which maintains a constant learning rate after warmup without any decay. Through experiments with 1B and 8B parameter models, we show that WSO consistently outperforms decay-based schedulers in terms of performance after SFT, even though decay-based schedulers may exhibit better performance after pre-training. The result also holds across different regimes with mid-training and over-training. Loss landscape analysis further reveals that decay-based schedulers lead models into sharper minima, whereas WSO preserves flatter minima that support adaptability. These findings indicate that applying LR decay to improve pre-training metrics may compromise downstream adaptability. Our work also provides practical guidance for training and model release strategies, highlighting that pre-training models with WSO enhances their adaptability for downstream tasks. --- ### SWE-QA-Pro: A Representative Benchmark and Scalable Training Recipe for Repository-Level Code Understanding Authors: Songcheng Cai, Zhiheng Lyu, Yuansheng Ni, Xiangchao Chen, Baichuan Zhou, Shenzhe Zhu, Yi Lu, Haozhe Wang, Chi Ruan, Benjamin Schneider, Weixu Zhang, Xiang Li et al. Published: 2026-03-17 Categories: cs.SE, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2603.16124v1 Agentic repository-level code understanding is essential for automating complex software engineering tasks, yet the field lacks reliable benchmarks. Existing evaluations often overlook the long tail topics and rely on popular repositories where Large Language Models (LLMs) can cheat via memorized knowledge. To address this, we introduce SWE-QA-Pro, a benchmark constructed from diverse, long-tail repositories with executable environments. We enforce topical balance via issue-driven clustering to cover under-represented task types and apply a rigorous difficulty calibration process: questions solvable by direct-answer baselines are filtered out. This results in a dataset where agentic workflows significantly outperform direct answering (e.g., a ~13-point gap for Claude Sonnet 4.5), confirming the necessity of agentic codebase exploration. Furthermore, to tackle the scarcity of training data for such complex behaviors, we propose a scalable synthetic data pipeline that powers a two-stage training recipe: Supervised Fine-Tuning (SFT) followed by Reinforcement Learning from AI Feedback (RLAIF). This approach allows small open models to learn efficient tool usage and reasoning. Empirically, a Qwen3-8B model trained with our recipe surpasses GPT-4o by 2.3 points on SWE-QA-Pro and substantially narrows the gap to state-of-the-art proprietary models, demonstrating both the validity of our evaluation and the effectiveness of our agentic training workflow. --- ### Functorial Neural Architectures from Higher Inductive Types Authors: Karen Sargsyan Published: 2026-03-17 Categories: cs.LG, cs.AI, math.AT, math.CT arXiv: https://arxiv.org/abs/2603.16123v2 Neural networks often learn the parts of a task but fail on novel combinations of those parts. We argue that this failure is architectural: a decoder generalizes compositionally only when it respects the algebraic laws of the task, i.e. when it descends from freely generated sequences to the quotient determined by those laws. We make this principle constructive by compiling Higher Inductive Type (HIT) specifications into neural architectures. Basepoints, path constructors, and 2-cells are mapped to base constraints, generator networks, structural concatenation, and learned homotopies. The resulting transport decoders are strict monoidal functors by construction: decoding a concatenated word is concatenation of independently generated loop segments. In contrast, we prove that softmax self-attention cannot simultaneously satisfy strict monoidal composition and descent to any non-trivial compositional quotient. Experiments on the torus, wedge of circles, and Klein bottle validate the predicted hierarchy: functorial decoders outperform non-functorial alternatives by $2$--$10\times$, and a learned 2-cell closes a $46\%$ error gap precisely on words exercising the Klein-bottle relation. These results suggest that compositional generalization should be enforced as functorial structure in the architecture, rather than learned from examples alone. --- ### Out-of-Distribution Object Detection in Street Scenes via Synthetic Outlier Exposure and Transfer Learning Authors: Sadia Ilyas, Annika Mütze, Klaus Friedrichs, Thomas Kurbiel, Matthias Rottmann Published: 2026-03-17 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.16122v1 Out-of-distribution (OOD) object detection is an important yet underexplored task. A reliable object detector should be able to handle OOD objects by localizing and correctly classifying them as OOD. However, a critical issue arises when such atypical objects are completely missed by the object detector and incorrectly treated as background. Existing OOD detection approaches in object detection often rely on complex architectures or auxiliary branches and typically do not provide a framework that treats in-distribution (ID) and OOD in a unified way. In this work, we address these limitations by enabling a single detector to detect OOD objects, that are otherwise silently overlooked, alongside ID objects. We present \textbf{SynOE-OD}, a \textbf{Syn}thetic \textbf{O}utlier-\textbf{E}xposure-based \textbf{O}bject \textbf{D}etection framework, that leverages strong generative models, like Stable Diffusion, and Open-Vocabulary Object Detectors (OVODs) to generate semantically meaningful, object-level data that serve as outliers during training. The generated data is used for transfer-learning to establish strong ID task performance and supplement detection models with OOD object detection robustness. Our approach achieves state-of-the-art average precision on an established OOD object detection benchmark, where OVODs, such as GroundingDINO, show limited zero-shot performance in detecting OOD objects in street-scenes. --- ### Language Models Don't Know What You Want: Evaluating Personalization in Deep Research Needs Real Users Authors: Nishant Balepur, Malachi Hamada, Varsha Kishore, Sergey Feldman, Amanpreet Singh, Pao Siangliulue, Joseph Chee Chang, Eunsol Choi, Jordan Lee Boyd-Graber, Aakanksha Naik Published: 2026-03-17 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.16120v2 Deep Research (DR) systems help researchers cope with ballooning publishing counts. Such tools synthesize scientific papers to answer research queries, but lack understanding of their users. We address this with MyScholarQA (MySQA), a personalized DR agent that: 1) infers a profile with a user's research interests; 2) proposes personalized actions for a user's input query; and 3) writes a multi-section report for the query that follows user-approved actions. We first test MySQA with NLP's standard protocol: we build a benchmark with synthetic users and LLM judges, where MySQA beats baselines in citation metrics and personalized action-following. However, we suspect this process does not cover all aspects of personalized DR users value, so we interview users in an online version of MySQA to unmask them. We reveal nine nuanced errors of personalized DR undetectable by our LLM judges, and we study qualitative feedback to form lessons for future DR design. In all, we argue for a pillar of personalization that easy-to-use LLM judges can lead NLP to overlook: real progress in personalization is only possible with real users. --- ### PathGLS: Evaluating Pathology Vision-Language Models without Ground Truth through Multi-Dimensional Consistency Authors: Minbing Chen, Zhu Meng, Fei Su Published: 2026-03-17 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.16113v1 Vision-Language Models (VLMs) offer significant potential in computational pathology by enabling interpretable image analysis, automated reporting, and scalable decision support. However, their widespread clinical adoption remains limited due to the absence of reliable, automated evaluation metrics capable of identifying subtle failures such as hallucinations. To address this gap, we propose PathGLS, a novel reference-free evaluation framework that assesses pathology VLMs across three dimensions: Grounding (fine-grained visual-text alignment), Logic (entailment graph consistency using Natural Language Inference), and Stability (output variance under adversarial visual-semantic perturbations). PathGLS supports both patch-level and whole-slide image (WSI)-level analysis, yielding a comprehensive trust score. Experiments on Quilt-1M, TCGA, REG2025, PathMMU and TCGA-Sarcoma datasets demonstrate the superiority of PathGLS. Specifically, on the Quilt-1M dataset, PathGLS reveals a steep sensitivity drop of 40.2% for hallucinated reports compared to only 2.1% for BERTScore. Moreover, validation against expert-defined clinical error hierarchies reveals that PathGLS achieves a strong Spearman's rank correlation of $ρ=0.71$ ($p < 0.0001$), significantly outperforming Large Language Model (LLM)-based approaches (Gemini 3.0 Pro: $ρ=0.39$, $p < 0.0001$). These results establish PathGLS as a robust reference-free metric. By directly quantifying hallucination rates and domain shift robustness, it serves as a reliable criterion for benchmarking VLMs on private clinical datasets and informing safe deployment. Code can be found at: https://github.com/My13ad/PathGLS --- ### ASDA: Automated Skill Distillation and Adaptation for Financial Reasoning Authors: Tik Yu Yim, Wenting Tan, Sum Yee Chan, Tak-Wah Lam, Siu Ming Yiu Published: 2026-03-17 Categories: cs.CL, cs.AI, cs.CE arXiv: https://arxiv.org/abs/2603.16112v2 Adapting large language models (LLMs) to specialized financial reasoning typically requires expensive fine-tuning that produces model-locked expertise. Training-free alternatives have emerged, yet our experiments show that leading methods (GEPA and ACE) achieve only marginal gains on the FAMMA financial reasoning benchmark, exposing the limits of unstructured text optimization for complex, multi-step domain reasoning. We introduce Automated Skill Distillation and Adaptation (ASDA), a framework that automatically generates structured skill artifacts through iterative error-corrective learning without modifying model weights. A teacher model analyzes a student model's failures on financial reasoning tasks, clusters errors by subfield and error type, and synthesizes skill files containing reasoning procedures, code templates, and worked examples, which are dynamically injected during inference. Evaluated on FAMMA, ASDA achieves up to +17.33% improvement on arithmetic reasoning and +5.95% on non-arithmetic reasoning, substantially outperforming all training-free baselines. The resulting skill artifacts are human-readable, version-controlled, and compatible with the Agent Skills open standard, offering any organization with a labeled domain dataset a practical and auditable path to domain adaptation without weight access or retraining. ---